Genetics of Congenital Arrhythmia Syndromes: The Challenge of Variant Interpretation

Abstract

Congenital arrhythmia syndromes are rare genetic disorders that can cause a high risk of sudden cardiac death. Expert panels have affirmed 15 genes that are linked to congenital arrhythmias. These genes mostly encode cardiac ion channel proteins or associated regulatory proteins that generate the cardiac action potential. Common genetic variation modulates the risk of rare variants and partially explains the incomplete penetrance of these disorders. As genetic testing becomes more prevalent, a major challenge is that most detected variants are annotated as variants of uncertain significance (VUS). This review will highlight emerging methods that are refining our understanding of arrhythmia genetics, including phenotype risk scores, large cohorts, in vitro functional assays, structural models, and computational predictions.

Introduction

Congenital arrhythmia syndromes include the long QT syndrome (LQTS), Brugada syndrome (BrS), catecholaminergic polymorphic ventricular tachycardia (CPVT), and short QT syndrome (SQTS). These rare syndromes (frequency of 1 in 2,000 to 1 in 10,000) are classically considered to be caused by rare, large-effect variants in arrhythmia genes. Atrial fibrillation (AFib) is the most common arrhythmia but is considered a common disease with a highly polygenic and environmental origin [1], and will largely not be discussed in this review. Arrhythmias can confer risk for sudden cardiac death (SCD) that partially can be prevented by drug therapy, avoidance of disease-specific triggers such as drugs or exercise, or implantation of cardioverter-defibrillators in high-risk individuals. Genetic testing is increasingly becoming a part of clinical care for arrhythmias. An expert statement from the American Heart Association [2] and a consensus statement from four international heart rhythm associations [3] both affirmed the utility of genetic testing in individuals with suspected congenital arrhythmia disorders or in close relatives of affected individuals, and outlined genetic testing considerations for specific conditions.

Studies over the last 30 years have identified the major genes underlying congenital arrhythmia syndromes: 1) genes encoding cardiac ion channels underlying the cardiac action potential, 2) genes affecting intracellular calcium handling, and 3) additional genes that regulate these processes (Figure 1). Dozens of gene-disease associations have been proposed, including over 20 genes for each of LQTS and BrS. Some purported gene-disease associations have only been supported by minimal evidence, but these genes were nevertheless added to expanding clinical genetic testing panels. In response, the Clinical Genomics Resource (ClinGen) systematically reassessed the validity of gene-disease associations, asserting that 15 genes have “definitive” or “strong” congenital arrhythmia associations (Table 1) [4-6].

Figure 1: Major genes associated with congenital arrhythmias.

A) Schematic of the mechanism of congenital arrhythmia genes. B) Major plasma membrane ion channels in the heart. C) Left: Diagram of a cardiomyocyte with major arrhythmia-associated proteins shown (not to scale). Right: Additional congenital arrhythmia-associated genes that are not plasma membrane-localized ion channels. B-C) Diseases affirmed by ClinGen at the definitive, strong, or moderate levels are in bold. Abbreviations: BrS—Brugada syndrome, LQT—long QT syndrome, DCM—dilated cardiomyopathy, SQT—short QT syndrome, JLNS—Jervell and Lange-Neilsen Syndrome, And.-Tawil— Andersen-Tawil syndrome, CPVT—catecholaminergic polymorphic ventricular tachycardia. RyR2 CRDS—RyR2 calcium release deficiency syndrome.

Table 1:

ClinGen reappraisal of genes associated with inherited arrhythmia syndromes

ClinGen Class	Long QT syndrome	Brugada syndrome	Catecholaminergic polymorphic VT	Short QT syndrome	Other
Definitive	KCNQ1, KCNH2, SCN5A, CALM1/2/3	SCN5A	RYR2, CASQ2(AR), TRDN, TECRL	KCNH2	CACNA1C (Timothy), KCNJ2 (Anderson-Tawil)
Strong	TRDN (AR)	—	—	KCNQ1, SLC4A3 *	KCNE1 (aLQTS), KCNE2 (aLQTS)
Moderate	CACNA1C	—	CASQ2(AD), CALM1/2/3	SLC4A3*, KCNJ2	—
Limited	CAV3, KCNE1, KCNJ2	—	—	—	—
Disputed/ No evidence	SNTA1, AKAP9, ANK2, KCNE2, KCNJ5, SCN4B	20 genes	ANK2, KCNJ2, PKP2, SCN5A	CACNA1C, CACNA2D1, CACNB2, SCN5A, SLC22A5 **	—

Adapted from ClinGen reappraisals of LQTS [4], BrS [5], and CPVT/SQTS [6].

^*The panel reviewing SLC4A3 was divided whether to classify it as having strong or moderate evidence.

^**This gene causes primary systemic carnitine deficiency which may mimic SQTS. Abbreviations: VT=ventricular tachycardia, AR=autosomal recessive, AD=autosomal dominant, aLQTS=acquired long QT syndrome. Timothy Syndrome and Anderson-Tawil Syndrome are characterized by long QT syndrome in conjunction with characteristic non-cardiac phenotypes. CALM1, CALM2, and CALM3 are 3 distinct genes that all encode an identical calmodulin protein.

Under the 2015 American College of Medical Genetics-Association for Molecular Pathology (ACMG-AMP) classification scheme, variants are classified as benign, likely benign, variant of uncertain significance (VUS), likely pathogenic, or pathogenic [7]. A growing challenge is to decipher which rare variants in arrhythmia genes are linked to arrhythmias. Unfortunately, a majority of discovered variants in arrhythmia genes are annotated as variant of uncertain significance (VUS) [8]. This review will highlight emerging high-throughput methods that use patient datasets, in vitro functional assays, computational predictors, and protein hotspot and 3D structural analyses to help resolve the VUS Problem. This review will also discuss a major emerging theme of arrhythmia genetics—that so-called “Mendelian” arrhythmia syndromes are also influenced by common genetic variants (variants with a minor allele frequency of at least 5%). These common genetic variants can be aggregated and quantified in a Polygenic Risk Score (PRS), which may have utility in arrhythmia prediction [9].

Genes associated with congenital arrhythmias

Long QT syndrome (LQTS):

LQTS is characterized by a prolongation of the QT interval on the electrocardiogram, with risk of the torsades de pointes arrhythmia and SCD. Heterozygous dominant LQTS variants are termed Romano-Ward syndrome. Variants that lead to loss of function of the repolarizing potassium currents I_Ks (encoded by KCNQ1 and KCNE1), I_Kr (KCNH2), or I_K1 (KCNJ2) can lead to LQTS [4,10]. In addition, gain of function variants in the SCN5A sodium channel or CACNA1C calcium channel genes cause excessive cation influx, leading to LQTS. Homozyous or compound heterozygous variants in KCNQ1 or KCNE1 can lead to Jervell and Lange-Neilsen Syndrome, characterized by marked QT prolongation and deafness [10,11].

Short QT syndrome (SQTS):

SQTS is a very rare syndrome characterized by a shortening of the QT (<340 ms) on the baseline ECG, with risk of subsequent arrhythmias. Gain of function variants in three potassium channel genes (KCNQ1, KCNH2, or KCNJ2) can cause SQTS, as well as loss of function variants in the anion exchanger SLC4A3 [6]. SLC22A5 loss of function causes primary systemic carnitine deficiency which may mimic SQTS [12].

Brugada syndrome (BrS):

BrS is characterized by ST-segment elevation in the right precordial leads on the electrocardiogram, with risk of SCD. Although over 20 genes have been linked with BrS, a ClinGen expert panel asserted that only dominant loss of function variants in the sodium channel gene SCN5A have definitive evidence, with the remaining genes disputed [5]. BrS appears to be more prevalent in individuals of Asian ancestry compared to other ancestries [13]. As is discussed below, BrS has a complex genetic basis with a large common genetic influence.

Catecholaminergic polymorphic ventricular tachycardia (CPVT):

CPVT is characterized by bidirectional or polymorphic ventricular tachycardia during exercise or adrenergic stimulation, which can lead to syncope or SCD. The most common cause of CPVT is dominant gain-of-function variants in the cardiac ryanodine receptor gene RYR2, which encodes the major channel involved in calcium-induced calcium release from the sarcoplasmic reticulum (SR) [14]. CPVT can also arise from dominant or recessive variants in the SR calcium-binding protein calsequestrin (encoded by CASQ2) [15] or recessive variants in the trans-2,3-enoyl-CoA reductase-like gene TECRL [16].

Other disorders:

Loss of function variants in RYR2 can cause the recently recognized RyR2 calcium release deficiency syndrome (CRDS), which (unlike CPVT) is not revealed by exercise testing, but also presents with ventricular arrhythmia and SCD [17]. A specific electrophysiological pacing protocol (a long-burst, long-pause, short-coupled stimulus) can induce ventricular arrhythmias in patients with CRDS [17]. Recessive null mutations in TRDN (triadin), a regulator of calcium release from the sarcomeric reticulum, can lead to triadin knockout syndrome (TKOS) [18]. In TKOS, young patients can present with T-wave inversions and QT prolongation or CPVT-like features, which can lead to recurrent ventricular arrhythmias even with treatment [18]. A high-penetrance form of LQTS presenting early in age can be caused by variants in the multi-functional calcium-binding protein calmodulin (encoded by 3 identical genes CALM1, CALM2, and CALM3) [19]. Variants in CALM1/2/3 can also lead to CPVT-like features. [19] A variety of additional congenital conduction disorders have been described, such as progressive cardiac conduction defect and sinus node dysfunction. In particular, SCN5A variants can give rise to a variety of conduction disorders including overlap syndromes with multiple different presentations [20].

Common genetic contributions to “Mendelian” arrhythmias

Classically, congenital arrhythmia syndromes were considered “Mendelian”, with rare large-effect variants conferring arrhythmia susceptibility. However, arrhythmia disorders were long-recognized to have incomplete penetrance—the phenomenon that only a fraction of individuals with a pathogenic variant demonstrate a phenotype. This led to the hypothesis that other genetic factors may be influencing arrhythmia risk, such as common genetic variation. The first genome-wide association study (GWAS) of a congenital arrhythmia syndrome, BrS, identified an unexpectedly large contribution of polygenic common loci near SCN5A-SCN10A and a transcription factor HEY2 [21]. The most recent BrS GWAS identified 21 significant associated common loci, including 8 independent loci near SCN5A and 6 loci near cardiac developmental transcription factors [22]. Genetic and functional studies identified MAPRE2, encoding a microtubule binding protein, as a regulator of Na_V1.5 trafficking and BrS risk [22].

Common genetic variants discovered through a GWAS can be aggregated in a weighted average called a polygenic risk score (PRS) [9]. PRS’s typically follow a normal distribution, and individuals in the upper tail of the distributions of PRS might be especially predisposed to disease (Figure 2). A GWAS of AFib in over a half million individuals identified 97 significant loci, the largest of which is near the transcription factor PITX2 [23]. A PRS for AFib was deployed in the UK Biobank and explained 4.7% of the risk of AFib [24].

Figure 2: Common polygenic variants impact arrhythmia risk.

A) Schematic of the relationship between variant allele frequency and effect size (impact on arrhythmia risk). Rare large-effect variants and common variants identified through genome-wide association studies are highlighted. B) Polygenic risk scores (PRS) in the general population can identify individuals with high PRS who may be at an elevated risk for arrythmia disorders. Comparison of controls and cases may reveal altered distributions of PRS, indicating the influence of common genetic variants on disease risk.

In recent years, multiple studies have attempted to decipher the role of common genetic variation in influencing the QT interval and LQTS. A GWAS of baseline QT interval in 109,000 individuals identified over 35 common loci that explained ~10% of the variance in QT interval [25]. A 61 SNP PRS derived from the baseline QT GWAS explained 5-10% of QT variation in the general population [26]. The baseline QT PRS was found to be enriched in LQTS cases [27] and cases of drug-induced QT prolongation and arrhythmia [28]. In approximately 20% of cases of LQTS, no variant is found; these genotype-negative individuals were especially enriched for high PRS’s [27]. Another analysis of genotype-positive LQTS patients found that a QT PRS only explained 2% of the variance in QT interval, about 5 times lower than in the general population [29]. This result mirrors a result for BrS, where a PRS had the largest impact on BrS risk in genotype-negative cases [30]. Finally, a 1.1 million SNP PRS for QT interval (including non-genome-wide significant SNPs) was recently developed. Analyses deploying this PRS in individuals with a prolonged QT interval revealed a mixture of rare, pathogenic variants (classic LQTS) and polygenic common risk (high PRS) [31].

In summary, congenital arrhythmia syndromes have a polygenic common genetic component and PRS’s may help predict individual risk for arrhythmias, especially in genotype-negative individuals. One limitation of PRS’s is that they have lower predictive accuracy outside the ancestry group in which they were developed. Large diverse GWAS’s that include non-European ancestry individuals are needed to develop accurate arrhythmia PRS’s across ancestry groups.

The VUS Problem and high-throughput functional analyses

A growing challenge in arrhythmia genetics is the “variant of uncertain significance (VUS) problem,” the issue that a large fraction of observed variants are VUS. This problem is especially acute for rare missense variants, which are plausible disease-causing variants [32] and are often classified as VUS. Over the past few years, the number of missense variants in arrhythmia-associated genes deposited in the ClinVar variant database has risen to over 5,000, and the percentage that are VUS or have conflicting classifications has risen as well (Figure 3A-B). In the past 10 years, 8.4% of arrhythmia variants have had a meaningful reclassification in ClinVar, and most reclassifications have been in the direction of diagnostic uncertainty, e.g. from pathogenic/likely pathogenic or benign/likely benign to VUS/conflicting [8]. Large genes pose an especially large problem. For example, 3% of individuals carry a rare variant in the very large (4,967 amino acid) gene RYR2, and most newly discovered RYR2 variants are annotated as VUS [14]. A likely reason that a large percentage of variants are VUS is that most variants are present at a very low frequency in the population and therefore have not have yet been the target of detailed patient or functional studies.

Figure 3: The variant of uncertain significance problem.

A,B) Prevalence of missense variants in arrhythmia genes in ClinVar over time. Yearly archives of the ClinVar dataset were downloaded and processed to include missense variants in the 15 ClinGen definitive or strong arrhythmia genes (Table 1). VUS=variant of uncertain significance, P/LP=pathogenic/likely pathogenic, B/LB=benign/likely benign. Panel B is identical to panel A but with a zoomed-in y-axis. C) High-throughput in vitro functional approaches can characterize variants at various levels of accuracy and scale. D) Schematic of computational and structural predictions of variant effect. E) Counts of P/LP, B/LB, or VUS/Conflicting variants by gene. The 15 ClinGen definitive or strong arrhythmia genes are shown. Clinvar data in E was obtained on February 2, 2022.

One approach to improved variant classification is to use improved patient phenotyping. Phenotype-enhanced variant classification, incorporating a clinical CPVT scorecard, resulted in lower rate of RYR2 VUS’s [33]. A similar approach incorporating type 1 LQT-specific clinical features also reduced the rate of KCNQ1 VUS’s [34].

Well-validated in vitro studies can be applied at a strong level in the ACMG classification scheme to aid variant classification [7], and several complementary in vitro functional approaches are being used (Figure 3C). Cardiac ion channel variants have classically been studied by low-throughput manual patch clamping. Automated planar patch clamping enables an increase in throughput, and has been recently deployed to study and reclassify dozens of variants in KCNQ1, KCNH2, and SCN5A [35-38]. These studies take place in non-cardiac cell lines, so an open question is how faithfully they reflect the function of channels in cardiomyocytes.

Human induced pluripotent stem cell-derived cardiomyocytes are a promising system to study arrhythmia variants and have been deployed in over 50 studies to date. iPSC-CMs can be studied from patient-derived lines with suspected arrhythmia conditions [39]. Alternately, variants can be introduced with CRISPR-Cas9 editing [40,41] or the endogenous gene can be knocked out and replaced with a variant gene [42,43]. iPSC-CMs are especially promising for the study of LQTS variants, because prolonged QT has a clear cellular correlate in iPSC-CMs—a prolonged action potential. iPSC-CMs also enable the study of non-coding variants, such as a cryptic deep intronic variant that affects KCNH2 splicing [44]. Challenges with iPSC-CMs remain, including challenges of throughput and cell variability. iPSC-CMs also have an immature phenotype with altered expression of several important arrhythmia-associated genes compared to CMs in the adult heart.

Deep mutational scanning is an emerging technology that can assay thousands of variants in a single multiplexed experiment. The process involves generating a library of variants, expressing that library in cells (1 variant per cell), and performing a selection using a cellular phenotype linked to the function of the protein of interest (e.g. fluorescence, cellular localization, or cell survival). High-throughput sequencing is used to identify functional and dysfunctional variants. Complete deep mutational scans of CALM1/2/3 [45] and KCNJ2 [46] have been published, as well as scans of partial regions of SCN5A and KCNH2 [47,48]. With deep mutational scans—as well as all the in vitro assays discussed above—a critical future question will be the throughput, cost, and accuracy of the assays in measuring variant function and disease risk (Figure 3C).

Computational and structural predictions

Computational predictors that incorporate multiple features, including three-dimensional protein structural data, may also help predict per-variant arrhythmia risk (Figure 3D). In the past several years, advances in cryo-electron microscopy has enabled the solving of structures of the channels encoded by RYR2 [49], KCNQ1 and calmodulin [50], KCNH2 [51], and SCN5A [52,53]. In addition, computational protein structure predictions have dramatically improved with the development of AlphaFold [54]. Improved structural knowledge enables the identification of hotspots or key functional residues that can be incorporated in variant effect predictors [14,55,56]. A recent study leveraged the homology of cardiac and non-cardiac voltage-gated sodium and calcium channels in a machine learning approach to separately predict loss or gain of function variants [57]. Other computational variant effect predictors have been developed for all genes [58,59] or specifically for arrhythmia-associated genes [60,61]. As computational and structural variant predictors continue to increase in accuracy, they can aid variant classification and improve understanding of underlying biological mechanisms.

Conclusions

In the past few years, expert panel reappraisals, high-throughput functional studies, and computational and structural predictors have helped decipher which genes and variants are linked to congenital arrhythmia syndromes. Increasingly large patient datasets are being used for genetic discovery and to clarify the contribution of common genetic variation. An improved understanding of genotype-phenotype relationships in congenital arrhythmias ultimately has promise to help identify at-risk patients and prevent morbidity and mortality associated with these disorders.